JPH0562797B2 - - Google Patents

Description

[Detailed Description of the Invention] [Summary] Image data distributed on a spherical surface is divided into equal numbers for each same latitude on the spherical surface and stored in a line, and image data corresponding to each latitude is stored in parallel. When performing peak extraction processing by comparing the density of the pixel of interest and neighboring pixels for the image data stored in the planar memory, the peak value for the image data is extracted from the peak value on the spherical surface. For each latitude,
A peak extraction method that prevents imbalance in processing results by validating every value indicating how many pixels in the planar memory one pixel at the latitude corresponds to.

[Industrial application field]

The present invention includes a planar memory that stores image data distributed on a spherical surface in equal numbers for each same latitude on the spherical surface in a line, and stores image data corresponding to each latitude in parallel. , relates to a peak extraction method in a spherical mapping device that performs peak extraction processing by comparing the density of a pixel of interest and neighboring pixels on image data stored in the planar memory.

In recent years, object recognition technology has been actively developed to enable machines to perform functions similar to those of the human eye. In such object recognition, a method is used in which the object is captured as an image, and information such as features of the object is extracted from this image to recognize the object. For this reason, the object is generally imaged by an imaging means such as a television camera, and the original image is captured.
It is necessary to perform electrical processing based on this image to extract information.

On the other hand, such original images often contain noise, blur, or blur, and it is necessary to remove these noise components and extract structural lines and the like that constitute the original image.

As a method for extracting such an image, an information extraction method using mapping is conventionally known. As an information extraction method using this mapping, a method using spherical mapping (Japanese Patent Laid-Open No. 60-218183, etc.) is known, but this method involves projecting the original image onto a spherical surface (spherical projection). A great circle centered on each point of the original image projected onto the image is drawn, and information is extracted from the mutual relationships between the drawn great circles.

By following these steps, we can obtain the normal vector of the plane that includes each point of the original image and the center of the spherical surface, and at the same time extract line segments, and it is also possible to reproduce broken or distorted line segments. Become.

Furthermore, by mapping the change information of the projection on the spherical surface to a single point on the spherical surface, the movement locus of the point-like object that moves in a straight line is extracted. Furthermore, the projections of a plurality of line segments on the spherical surface are each mapped to one turn on the spherical surface, and then by mapping these mapped points, the original intersection points of the line segments are reproduced.

By the way, in order to handle the mapping function of a great circle corresponding to each point on the spherical surface, it is necessary to expand it into a two-dimensional memory.

[Conventional technology]

FIG. 6 is an explanatory diagram of a method for storing data represented by latitude and longitude coordinates on a spherical surface in association with vertical and horizontal coordinates on a two-dimensional memory on a plane.

As shown in Figure a, a point 2 on a spherical surface 1 is expressed by the latitude r from the equator 3 (divided into M equal parts) and the angle θ from the reference line 4 (divided into N equal parts), and its memory area 5 is
As shown in Figure b, the vertical axis is r and the horizontal axis is θ.

That is, one pixel near the equator on the sphere is stored in correspondence with one pixel on the two-dimensional memory, and one pixel near the north pole on the sphere is stored in correspondence with multiple pixels on the two-dimensional memory. Ru.

[Problem that the invention seeks to solve]

However, if you simply divide the latitude and longitude into equal numbers, quantize each, and develop the sphere into a two-dimensional memory, there is a problem that the area of the area occupied by the same angle is different between the lower and higher latitude areas of the sphere. occurs.

In particular, near the North Pole, the resolution becomes excessive, and the peak value becomes extremely broad on the grid-like memory, and the top often appears to extend over multiple pixels on the two-dimensional memory.

Therefore, in conventional peak extraction, for example, a 3×3 matrix window is used to determine whether the density value of the center pixel of interest is higher than the density values of the surrounding pixels, and the top All the pixels in the area are extracted.

In this case, when spherical mapping is performed on the image data on the two-dimensional memory having such broad peak values, the number of great circles corresponding to the number of peaks is on the next two-dimensional mapping memory. On the other hand, in areas where the resolution is not excessive, a great circle is added only once.

Therefore, if some parts have excessive resolution and some parts do not, there is a problem that information imbalance occurs when the next stage of mapping is performed.

In view of the above-mentioned conventional problems, an object of the present invention is to
Even if the resolution of the length of each pixel on the spherical surface differs depending on the latitude, the spherical surface is expanded into a two-dimensional memory by dividing the latitude and longitude into equal numbers and quantizing each. An object of the present invention is to provide a peak extraction method that can prevent information imbalance from occurring during mapping.

[Means for solving problems]

An object of the present invention is to provide a planar memory that stores image data distributed on a spherical surface in equal numbers for each same latitude on the spherical surface in a line, and stores image data corresponding to each latitude in parallel. In a peak extraction method in a spherical mapping device, which performs peak extraction processing by comparing the density of a pixel of interest and neighboring pixels for image data stored in the planar memory, , find a value indicating how many pixels in the planar memory one pixel at the latitude corresponds to, and when performing peak extraction processing of the image data at each latitude, make the peak value for the image data valid every other value. This is achieved by a peak extraction method characterized by the following.

〔Example〕

Embodiments of the peak extraction method according to the present invention will be described in detail below with reference to the drawings.

In the present invention, the spherical surface is quantized as shown in Figure 3, and as shown in Figure 4, one pixel near the north pole on the spherical imaging surface corresponds to multiple physical pixels on the two-dimensional memory. Therefore, when performing peak extraction processing using a matrix, for each latitude, find out how many pixels on the spherical surface corresponds to one pixel on the two-dimensional memory, and then By extracting image data and performing peak extraction processing, there will be no imbalance of information during the next stage of mapping.

Below, a method for determining how many pixels on the two-dimensional memory corresponds to one pixel on the spherical surface will be explained.

First, as shown in FIG. 5, the size of a pixel on a spherical surface is assumed to be Δβ ₀ .

Also, as shown in Figure 6, the quantization width on the two-dimensional memory is ^d Δβ ₀ in the latitude direction and ^d Δβ 0 in the longitude direction.
Let Δα be ₉₀ .

Here, the distance between a pixel of interest on the spherical surface and a pixel adjacent in the longitudinal direction is ^r Δθ = Δβ ₀ /sinβ on the two-dimensional memory, where β is the angle when the north pole is set to 0 degrees. It is. is required.

In other words, at the north pole of the sphere, ^r Δθ = ∞, so there are no adjacent pixels on the two-dimensional memory, and near the equator of the sphere, ^r Δθ = 1, and the pixels on the sphere and the two-dimensional memory There is a one-to-one correspondence with the pixels of .

Furthermore, the number of pixels on the two-dimensional memory that one pixel on the spherical surface corresponds to is expressed as Δm, and is determined by Δm= ^r Δθ/ ^d Δα ₉₀ .

Therefore, when calculating this value Δm for each latitude and performing peak extraction using a 3 x 3 matrix, for example, perform peak extraction processing for each point separated by Δm, or perform peak extraction using a 3 x 3 matrix. After the processing, only the peak values at points separated by Δm are set as the true peak values, so that information imbalance does not occur during the next mapping.

FIG. 1 is a block diagram showing the configuration of an image processing apparatus using the peak extraction method of the present invention.

In the figure, 30 is a microprocessor (hereinafter referred to as
31 is a spherical camera, 32 is an original image memory, 33 is a matrix calculation unit, 34 is a Δm value storage table, 35 is an image processing result storage memory,
36 is a mapping processor.

The MPU 30 includes a spherical camera 31, an original image memory 32,
It controls the matrix calculation unit 33 etc.
The spherical camera 31 captures an image of the object and outputs the image data obtained through contour extraction processing in polar coordinate (r, θ) format, and the original image memory 32 captures the image data output from the spherical camera 31 as shown in FIG. Image data in polar coordinate format is stored in a grid-like manner, with storage areas divided into N pieces in the longitude direction and M pieces in the latitude direction, and areas near the equator are stored in one-to-one correspondence. However, the area near the north pole is stored in one-to-multiple pixel correspondence, and the matrix calculation unit 33
For example, performs a 3×3 matrix operation,
One that performs peak extraction processing as described below.
The Δm value storage table 34 stores the values obtained for each latitude as described above, and the image processing result storage memory 35 stores the values determined for each latitude.
3, the mapping processor 36 generates a mapping function (great circle) for the image data stored in the original image memory 32,
The image data is written to an internal mapping memory (which stores image data in the same storage format as the original image memory 32).

FIG. 2 is a diagram illustrating a method of peak extraction processing using a 3×3 matrix.

In this peak extraction process, a 3×3 matrix is set up with a pixel of interest E and four pixels B, D, F, and H surrounding this pixel E, and the pixel of interest E and each of the surrounding pixels B to H in this matrix are This is a process in which the values of the pixel of interest E are compared, and if the value of the pixel of interest E is greater than or equal to the values of the surrounding pixels B to H, the pixel of interest E is determined to be a peak point.

The operation of the configuration described above will be explained.

First, the spherical camera 31 images the object under the control of the MPU 30, outputs image data subjected to contour extraction processing in polar coordinate (r, θ) format,
The MPU 30 stores image data (coordinate positions and density values) from the spherical camera 31 in the corresponding storage area of the original image memory 32.

Then, once the image data for one screen has been stored in the original image memory 32, the MPU 30 issues a start command to the mapping processor 36.

Upon receiving the activation command, the mapping processor 36 sequentially generates mapping functions (great circle information) for the pixels stored in the original image memory 32, and writes and stores the values in the internal mapping memory.

Then, once the great circle information corresponding to each pixel in the original image memory 32 is stored in the mapping memory in the mapping processor, the MPU 30 reads the histogram value for each pixel from the mapping memory in the mapping processor 36, and performs matrix calculation. The peak extraction process is performed using the unit 33.

This peak extraction process is performed under the control of the MPU 30, and first, the mapping processor 36
The image data at the coordinates (r _M , θ ₀ ') of the mapping memory in the map memory is extracted as the value of the pixel of interest E, and the image data at the coordinates adjacent to it in the vertical and horizontal directions are extracted as the value of the surrounding pixels D to H. Extract it as a value,
These values are converted to 3×3 in the matrix calculation unit 33.
Set in matrix.

The matrix calculation section 33 performs peak extraction processing by performing the above-described processing based on the set values, and stores the processing results in the image processing result storage memory 35.

Next, the MPU 30 reads out the image data on the mapping memory in the mapping processor at a position corresponding to the vicinity of the North Pole (latitude value = r _M ).
A predetermined value Δm is read out from the Δm storage table 34 in accordance with this latitude value r _M.

Then, the coordinates obtained by adding this value Δm to the coordinates of the image data processed last time (r _M , θ ₀ +Δm)
The image data at the position is read out from the mapping memory in the mapping processor to be used as the value of the next pixel of interest E.

From this point on, the operations in the above sections are repeated in sequence.

In this embodiment, a case has been described in which the calculation position using the 3×3 matrix is set at every value of Δm, but the present invention is not limited to this, and all the positions on the mapping memory in the mapping processor are After performing peak extraction processing on the pixels, peak values may be extracted at intervals of Δm.

Further, in this embodiment, the case where there is one mapping processor has been described, but the present invention is not limited to this, and the number can be increased as appropriate depending on the type of information to be extracted. In this case, it goes without saying that the peak extraction method of the present invention may be applied to the mapping processing results of each mapping processor and stored in separate mapping processing result storage memories.

〔Effect of the invention〕

As explained above, according to the present invention, latitude,
Even if the longitude is divided into equal numbers and quantized and the spherical surface is developed into a two-dimensional memory, and the length resolution of each pixel on the spherical surface is different for each latitude,
When performing peak extraction processing using an n×n matrix, for each latitude, determine how many pixels on the spherical surface corresponds to one pixel on the two-dimensional memory, and then By extracting image data and performing peak extraction processing, there will be no imbalance of information during the next stage of mapping.

[Brief explanation of the drawing]

Figure 1 is a block diagram of the configuration of the image processing device;
The figure is an explanatory diagram of a 3x3 matrix, Figure 3 is a diagram showing the quantization method for the spherical surface, Figure 4 is a diagram showing the relationship between pixels on the spherical surface near the north pole and physical pixels, and Figure 5 is a diagram of the imaging plane. FIG. 6 is a diagram showing a state in which resolution is expressed on a spherical surface, and is an explanatory diagram of a storage method in a two-dimensional memory. In the figure, 30 is an MPU, 31 is a spherical camera,
32 is an original image memory, 33 is a matrix calculation unit, 3
4 is a Δm value storage table, 35 is an image processing result storage memory, and 36 is a mapping processor.

Claims (1)

[Scope of Claims] 1. A planar shape in which image data distributed on a spherical surface is divided into equal numbers at the same latitude on the spherical surface and stored in a line, and image data corresponding to each latitude is stored in parallel. In a peak extraction method in a spherical mapping device that is equipped with a memory and performs peak extraction processing by comparing the density of a pixel of interest and neighboring pixels for image data stored in the planar memory, each latitude on a spherical surface is Then, a value indicating how many pixels in the planar memory one pixel at the latitude corresponds to is obtained, and when peak extraction processing is performed on the image data at each latitude, the peak value for the image data is validated every other value. A peak extraction method characterized by: